bioemerl

25 ms·

331.
by bioemerl 4y ago
That's because these things are way more than word prediction. Their job is 100 percent to predict the next word, but saying that hides what you have to do in order to predict the next word In a factual database? You have to repeat fa
332.
by bioemerl 4y ago
They don't have to, they only have to be relative equals. It's possible chat GPT is a "dreadnaught moment"
333.
by bioemerl 4y ago
> When Sundar reacts fast enough Their response wasn't fast enough. They should have been experimenting with this stuff in public for years now, but they've kept it under cover. This was a deflated promise from Google. They s
334.
by bioemerl 4y ago
Problem you run into is that the screens need a pretty fair amount of power and phones just can't deliver it. USB-C can't deliver it either.
335.
by bioemerl 4y ago
I'm not really talking about LLMs specifically. I'm talking about AI as a field. Although neural networks are capable of any form of reasoning or understanding and human language is a very strongly diverse field so I do believe L
336.
by bioemerl 4y ago
Because we are a living example of what is eventually possible with "artificial" intelligence. If we exist, it can at least match us.
337.
by bioemerl 4y ago
I don't think so. The end point of that curve is models with human like reasoning or even superhuman ability. I think we may be on the S part of the curve, but thanks to processing concerns we will be in for veeeeeerrryyyy gentle slop
338.
by bioemerl 4y ago
> I'll start playing with ChatGPT and the like when I can run them on my own hardware. Koboldai
339.
by bioemerl 4y ago
What if I provide it for free and a business starts using it?
340.
by bioemerl 4y ago
It's like apple. Chick filet managed to market to the masses while feeling upscale.
341.
by bioemerl 4y ago
Man, that's already how a decent percent of recipes work. They come off cans and exist in part to sell products.
342.
by bioemerl 4y ago
An LLM could theoretically create a model with which to understand chess and predict a next move, you just need to adjust the training data and train the model until that behavior appears. The expressiveness of language lets this be true of
343.
by bioemerl 4y ago
We don't judge AI by their ability to produce language, we judge them by their conference and ability to respond intelligently, to give us information we can use.
344.
by bioemerl 4y ago
Exactly. Since language is a compressed and transmittable result of our thought, to predict text as accurately as possible requires you do the same. A model with understanding of the human mind will outperform one without. > Being able
345.
by bioemerl 4y ago
I do believe if we are going to get AGI without some random revolutionary breakthrough, to achieve it iteratively, It's going to come through language models. Think about it. What's the most expressive medium we have which is also
346.
by bioemerl 4y ago
Because we are talking about uses not involving copy or distribution of a work. The use of the images to train an AI - such uses have never been protected.
347.
by bioemerl 4y ago
All of these involve the reproduction of the thing. Training an AI model isn't responding the art. I can also*do* all of those things, I just can't sell it. That's what copyright is. Copy right. Not "use for training an
348.
by bioemerl 4y ago
Such as? Every example to date I've seen regards the copying or distribution of a product, not its use once it's in someone's hands.
349.
by bioemerl 4y ago
Essential is probably not right, this is true. But I'm confident I'll make a better product with telemetry vs without.
350.
by bioemerl 4y ago
Oh shit, I missed it as well.
351.
by bioemerl 4y ago
Mandatory with an open license might shoot yourself in the foot when people fork your code. For me, the harshest you can go is to have telemetry and not prompt for the setting on start with the user opted in by default. Ideally, you ask on
352.
by bioemerl 4y ago
I think it's somewhat silly to fly blind on the assumption that your taste is better than any real world observations you can make. Especially if you haven't had the chance to develop an intuition yet and are new to the field. Wi
353.
by bioemerl 4y ago
That is fine, but in this case telemetry trades you (and other more hardline users) as a user for all the extra users you gain from instant crash reports, quick feedback, and generally better productivity. I would never personally make that
354.
by bioemerl 4y ago
Really most of your post here is answered by the later fifth or so of the post you're responding to. I would reread that part. Models output very similar work to training data at times, even identical in some cases, but the vast majori
355.
by bioemerl 4y ago
Honestly, this may be unpopular with hacker news, but just add your own telemetry. If people don't like it they can turn it off, and telemetry is essential for a good product. Do let people turn it off though please.
356.
by bioemerl 4y ago
The alternative is not this. The alternative is you not being able to get data while the companies can.
357.
by bioemerl 4y ago
No need to be cynical or expect it to be slow. Multiple companies will leap at this precident if it is set. They've attempted to do it in the past a few times, but right to repair laws and such have kept them away. Imagine Nintendo b
358.
by bioemerl 4y ago
> it's not a totally different ball game without really providing any explanation of why. Because of I buy a book from a store, you now no longer control that book. I can read it. I can use it for just about any purpose I want to.
359.
by bioemerl 4y ago
In this case it should be pretty easily resolved, there's never been a right under copyright to control who sees or uses your work. You have the right to control who can copy and redistribute your work, which is a totally different bal
360.
by bioemerl 4y ago
You're kind of arguing against a straw man. The people who don't like these models don't like the ones that are posed to replace artists. They are not really worried about stuff like clip groups images together. I do believe

More ›